Quantitative modeling assesses AI infrastructure risks through several methods. A Monte Carlo simulation of 100,000 trajectories over 24 months, using expansion and disillusionment drifts of 15% and −25%, volatilities of 20% and 45%, a monthly jump probability of 3%, and normally distributed jumps with mean −35% and standard deviation 10%, yields cumulative losses at the 50th, 90th, 99th, and 99.9th percentiles of −8.2%, −32.5%, −58.7%, and −74.3%, respectively.
The 99% VaR indicates potential losses exceeding half the capital invested in AI infrastructure. Merton’s structural model estimates OpenAI’s distress probability at 14.7%, based on an implied IPO valuation of $852 billion, contractual obligations and debt of $600 billion, asset volatility of 50%, a five-year horizon, and a 4% risk-free rate. Systemic risk estimates show Nvidia’s monthly 5% VaR worsening from −8.5% to a conditional CoVaR of −18% during OpenAI distress, a ΔCoVaR of −9.5 percentage points, indicating significant, nonlinear contagion.
Nvidia’s GARCH-X model uses parameters ω = 0.0004, α = 0.08, β = 0.88, and γ = 0.03, capturing volatility clustering and sensitivity to credit news; Heston parameters κ = 2.0, θ = 0.04, ξ = 0.60, and ρ = −0.50 indicate a pronounced volatility skew and demand for put protection.
A Markov-switching model distinguishes Boom and Stress regimes, with quarterly transition probabilities of 92% and 85% persistence, a current Stress probability of 38% according to Aladdin 12.4, and a 22% probability of exiting the Bubble regime within 30 days.
A CCO Monte Carlo simulation of 100,000 trajectories across 50 GPU/neocloud exposures, assuming 10% marginal default probability, 0.35 default correlation, and 55% loss given default, estimates expected losses and 95% VaR at 32% and 72% for equity (0–10%), 8.5% and 24% for mezzanine (10–30%), and 1.1% and 4.6% for senior (30–100%) tranches; high correlation leaves senior debt exposed to tail losses.
Finally, game theory predicts a Nash equilibrium of continued investment by both hyperscalers during Boom and mutual cuts during Stress. Meta, led by Mark Zuckerberg, is identified as the rational first mover because of its dual-class control and the market’s favorable response to its 2022 capital expenditure cuts, whereas Oracle cannot reduce spending without triggering a credit event.
